Consumer demand identification is the core problem of modern business intelligence, and traditional methods are difficult to effectively integrate heterogeneous data from multiple sources. With the explosive growth of data scale on e-commerce platforms, how to accurately recognize consumer demand from multimodal data such as massive reviews and product information has become a key challenge. Existing demand identification methods are deficient in feature extraction and classification accuracy, and lack the ability to deeply integrate multimodal data, so there is an urgent need for more accurate and efficient technical means to realize the accurate identification of consumer demand. Purpose: Aiming at the problems of low accuracy and limited data fusion capability of traditional consumer demand identification methods, we propose a consumer demand identification method based on K nearest neighbor algorithm and multimodal data fusion. Methods: Web crawler technology is used to obtain 11,896 Jingdong Mall review data, and 10,861 valid texts are obtained after preprocessing; Jieba partitioning and TF-IDF algorithm are used to extract key features, and Particle Swarm Optimized K Nearest Neighbors (PSO-KNN) classification model is constructed to integrate multimodal data, such as semantic analysis and commodity information, for demand identification. Results: the proportion of positive consumer evaluations higher than 0.5 is 86.7%, the price of goods is mainly concentrated in the range of 10-20 yuan, and the goods with PH value distribution in the range of 3.5-4.5 are the most popular; compared with the traditional KNN and SVM algorithms, the average absolute percentage errors of PSO-KNN model are reduced by 1.87% and 3.18%, respectively. Conclusion: The proposed method effectively improves the accuracy of consumer demand identification and provides scientific support for enterprise precision marketing and product optimization decision-making.
